Turn on new tiling by default

[eellison]

Stack from ghstack (oldest at bottom):

• enable test #155342
• -> Turn on new tiling by default #154768

Turning on in fbcode to come. Also updates max_tiles to have a default value of None. The existing tiling logic doesn't really handle max_tiles=3 well, but we do in the new tiling logic, so we default to 3 in the new logic and 2 elsewhere unless max_tiles has been explicitly set.

TB runners have been very unstable recently (do we need to bump batch size ?) but e.g. for a recent torchbench inference run we had 15 models with a lower execution time (i.g. green) and 2 models with higher (i.e.. red)

I am doing another run and will update here.

Dynamic shapes is not yet turned on because there are a lot of fixes to be done in splitting that don't work yet.. See:

(Pdb) p expr
((s25*s85)//32)
(Pdb) p FloorDiv(expr, expr)
((s25*s85)//(32*(((s25*s85)//32))))

and also - unbacked shape is not multiple of itself.

cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @chenyang78 @kadeng @muchulee8 @amjames @chauhang @aakhundov

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/154768

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

⏳ 1 Pending, 1 Unrelated Failure

As of commit 744b0fb with merge base 4405dc1 ():

FLAKY - The following job failed but was likely due to flakiness present on trunk:

• pull / linux-focal-py3.9-clang10 / test (dynamo_wrapped, 1, 3, linux.2xlarge) (gh) (similar failure)
  'test/test_reductions.py::TestReductionsCPU::test_sum_all_cpu_float64'

UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:

• ⏳ pull / linux-jammy-py3-clang12-executorch / build (gh) (#150261)

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[eellison]

@pytorchbot merge -f "timing out"

[pytorchmergebot]

Can't merge closed PR #154768

[malfet]

@pytorchbot revert -m "Looks like it broke inductor CPU, see https://hud.pytorch.org/hud/pytorch/pytorch/231eb9902ba78a4ef70203243058f3c7c0ced15d/1?per_page=50&name_filter=inductor-triton-cpu&mergeEphemeralLF=true" -c nosignal

[pytorchmergebot]

@pytorchbot successfully started a revert job. Check the current status here.
Questions? Feedback? Please reach out to the PyTorch DevX Team

[pytorchmergebot]

@eellison your PR has been successfully reverted.

[eellison]

could we get inductor cpu in tests ?

[eellison]

@pytorchbot merge

[pytorchmergebot]

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here